feat(P3): implement additional rates + some fixes

[haakon-e]

Purpose: Integrate the P3 ice scheme with the two-moment microphysics

This branch makes the P3 ice category usable as a coupled component of the two-moment scheme, rather than a standalone solver. The motivation is to expose P3 ice processes (collisions, melting, nucleation, sedimentation) through the shared bulk-tendencies entry point so that a host model can drive 2M liquid and P3 ice consistently and performantly.

Rationale for the principal changes:

• Explicit P3State. Ice shape/threshold properties were previously recomputed ad hoc at each call site. Centralising them in a single, type-stable state object removes redundant work, makes the per-process interfaces uniform, and is a prerequisite for the GPU-ready paths below.
• Rates without time-step limiting. Melting and nucleation now return true instantaneous tendencies instead of pre-capping to the available mass/number over dt. Capping is the integrator's responsibility; doing it inside the process broke composability across sub-steps and conflated the physics with the numerics.
• 2M+P3 collision path and BMT coupling. A type-stable, allocation-free liquid–ice collision path and the bulk-tendencies entry point are what let P3 ice run on GPU at acceptable cost and interoperate with the existing 2M tendencies.
• Frostenberg INP nucleation. Adds an immersion-freezing parameterisation needed for realistic heterogeneous ice formation.

Tests and reference values were updated to reflect the new (uncapped) rates and the refactored APIs; numerical smoke references that shifted were re-derived and verified.

[haakon-e]

This change is part of the following stack:

• Update the CI to Julia v1.12 #704
  • ci: update tests, JuliaFormatter v2 #706
    • rft: improve parameter constructors #700
      • fix: 2M tendencies, params #694
        • feat(P3): implement additional rates + some fixes #681 ◀

_{Change managed by git-spice.}

[haakon-e]

comments to assist review.

[haakon-e]

dividing by 900 here is related to my note to self in src/parameters/MicrophysicsP3.jl. Need to check which ice densities I need.

[trontrytel]

Why is this in Utilities and not P3 code?

[haakon-e]

I kept them here since they're very simple wrappers so it felt natural to have them near the "actual" code. Then I didn't want sgs_weight_function and _regularised_ratio in P3 code, so I put them here.

Happy to move them